Cooperative and collaborative strategies for transmission in wireless sensor networks enable transmission range restricted nodes to reach distant receivers by superimposing transmission signals. This addresses an important practical problem of wireless sensor networks. In this proposal we extend this strategy by emergent properties: We establish a method to adapt the collaborative emergent optimisation process by a) remembering previous behaviour from similar situations, b) using this information to adapt the current optimisation run by using randomised and feedback-based approaches to determine an optimally pre-synchronised set of nodes for transmission and c) optimising and learning observed optimisation behaviour for the random process, which is better than the behaviour we had in memory so far. Using feedback information is a natural and intuitive approach to adapt to the the scenario’s dynamics without the requirement for external intervention. Our approach will therefore show both emergent and self-organisation properties. We will demonstrate the suitability of the method by implementing and deploying a sensor network in an office setting. The demonstration will show how to globally minimise and equalise the energy used for collaborative transmission.

Research Topics

• Developing an emergent radio for collaborative transmission
  In collaborative transmission, a set of nodes superimposes their transmit signals in order to collaboratively reach a remote receiver. In an iterative process the transmit signals are synchronised at a receiver location by phase-shifting the base band signals of distinct transmitters. This synchronisation process is applied in advance of each single transmission process, since the signals are synchronised only on a specific receiver location.
  However, since we assume sensor nodes to have a fixed location, the optimisation scenario is similar for each optimisation process as the number and location of nodes in the environment as well as environmental noise remain mostly constant. We claim that for this reason the optimisation process is also similar for each single optimisation. In particular, the optimisation speed at a given advance in the optimisation process, the channel quality of an optimum solution and the ratio of the distance to the optimum synchronisation and the current channel quality are expected to evolve similar in a specific environment. In an emergent process the sensor network is therefore able to adapt to the environmental situation in a self-improving manner.
  Adaptation of the network is feasible when the global optimisation approach for collaborative transmission is spatially divided between a receiver and the network as, for example, proposed in \citeown{4019,4020}. In these evolutionary approaches, the receiver is responsible for generating the channel quality feedback (the fitness function) while the nodes are capable of adapting the mutation and crossover operators. Since the optimisation speed and process is likely similar for each single optimisation run in a given scenario, we propose to adapt crossover and mutation parameters to this process. One possible solution we will consider is the consideration of an optimisation table for each node in which fitness values and optimisation speed (distance between successive fitness values) are related to crossover and mutation operators applied.
  Since crossover is more useful at the beginning of the optimisation and a small mutation probability can slow down the optimisation process near the optimum, knowledge about prior and current optimisation process is valuable to reduce the optimisation time in a given scenario.
  A further parameter that might foster emergent operation is the transmission power. Since in an environment with a high density of transmitters, the noise power is highly dependent on the transmission power of nodes, we also expect a potential for emergent behaviour in adapting the transmission power to the environmental setting.
  We propose for the nodes to utilise optimisation tables to update and look up proper optimisation parameters during the optimisation process. Optimisation tables are updated and improved as new or better configurations are observed. Collaborative transmission is then becoming more environmental adaptive and emergent.
  We will elaborate this emergent process in analytic studies and field tests in small or medium sized networks of sensor nodes.
• How to determine the optimum count of transmitting nodes
  In collaborative transmission approaches, nodes combine their transmit signals in order to increase the transmission range. The possible transmission range is dependent on the count of nodes that collaborate. With more nodes participating, the possible gain in transmission range is higher. However, the expected optimisation time is also increased as the number of transmit signals that are to be synchronised increases. Together with the signal count and the optimisation time, the overall energy consumption in the network increases. It is therefore desired to determine in advance of transmission the node count for collaborative transmission that optimises the overall energy consumption in the network so that the receiver node is still reached by the synchronised superimposed signal. In this manner, collaborative transmission is enhanced by a self-optimisation capability since the overall lifetime and connectivity of the sensor network is improved.
  In [Krohn07] an approach to estimate the number of transmitters in a small network of unsynchronised nodes was detailed. We will improve this approach so that it can also be applied for networks of great size. We will determine the approximation quality of this approach analytically. It was derived in [Krohn07] that the estimation quality can be improved by multiplying the transmit signal with an appropriate pseudo-noise sequence. The project will derive pseudo noise sequences that improve the estimation of the count of transmitting nodes.
  The approach will furthermore be evaluated in simulations (e.g. Matlab) and in medium sized networks of real sensor networks (e.g. uParts [Beigl2006, Decker2005, Beigl2005a])
  A related problem is the determination of the distance between the network and the receiver. While the round trip time, which is implicitly provided by the collaborative transmission approach might a rough estimate on this measure, complex scenarios require more ambiguous approaches. In scenarios in which a receiver is located in between sensor nodes (e.g. when sensors are distributed in a room at walls, floor and ceiling) rather than on one side of the network, The round trip time to several nodes might differ greatly.
• Determination of a set of pre-synchronised nodes
  While the restriction of the number of nodes participating in collaborative transmission will already decrease the synchronisation time and energy consumption for collaborative transmission, the amount of pre-synchronisation of transmit nodes also impacts these parameters.
  Assume two sufficiently different receiver locations A and B and an arbitrary set of nodes form the sensor network for which the base band frequency of the transmit signals is synchronised at receiver location A. the same set of nodes is then most likely unsynchronised at location of receiver B since the path lengths of the transmit signals differ for this alternative location. There might, however, exist a second, not necessarily disjoint set of nodes that is better pre-synchronised for a receiver at location B.
  The optimisation task is therefore to determine for a given receiver location that set of nodes that constitutes the best pre-synchronisation of transmit signals since this node set minimises the expected optimisation time and consequently the expected energy consumption. This approach is capable of further improving the ability for self-optimisation of a collaborative transmission approach since the optimisation speed, the energy consumption and consequently the lifetime and connectivity of a collaboratively transmitting sensor network is improved by this approach.
  We will study various randomised approaches to this problem that might include an initialisation phase in which the receiver might estimate the channel quality of random sets of nodes. A straightforward approach would, for example, consider iterations in which random sets of nodes collaboratively transmit and thereby provide a sample on their amount of presynchronisation. If this should be possible, we will also try to derive a deterministic approach to determine the optimum share of nodes that participate tin the collaborative transmission process.
  A problem expected with this approach is the ability to sustain network connectivity. Since the energy level of nodes is not considered in the decision on the node set, some nodes might participate disproportionally often.

Project

The project “Emergent Radio: Emergent strategies to optimise collaborative transmission schemes” is part of the DFG – priority programme Organic Computing (DFG SPP 1183).

Related Publications

• Albert Krohn: Superimposed Radio Signals for Wireless Sensor Network, Technical University of Braunschweig, Januar 2007 (Krohn07)

• Reisch, Mathias, cand.-wirt.-inf.: Particle Location System, Studienarbeit, 2008 (D.09.3002)